Inside the operating room of Baylor College of Medicine, researchers placed the finest recording electrodes directly into the brains of four bilingual patients—and discovered that the human mind keeps Spanish and English tangled together at the deepest level.
The finding matters because bilingualism has long puzzled neuroscientists. How do people speak two languages fluently without constantly confusing them? How can a bilingual person translate a thought between languages in milliseconds? For decades, the prevailing theory suggested the brain maintains separate neural filing systems for each language, like two distinct libraries. What neurosurgeon Dr. Sameer Sheth and his team at Baylor found instead was something far more elegant: a shared neural architecture that organizes meaning in the same geometric space, regardless of which language is being used.
Using ultra-high-resolution microelectrodes and Neuropixels probes, the researchers recorded the activity of hundreds of individual neurons in the hippocampus while four fully bilingual English-Spanish speakers listened to stories, read phrases aloud, and had spontaneous conversations in both languages. The key discovery was counterintuitive. Although most individual neurons fired differently when a person heard English versus Spanish, the overall pattern—what neuroscientists call "semantic geometry"—remained remarkably consistent across languages. The word "cat" and its Spanish equivalent "gato" occupied similar positions in neural space. Concepts with related meanings, like "cat" and "dog," stayed in the same relative relationship to each other whether someone was thinking in English or Spanish.
"Different languages appear to access a shared conceptual map rather than creating entirely separate representations of the world," Sheth explained. The researchers found a small population of neurons that responded identically to translation-equivalent words like "earth" and "tierra"—cells they termed "cross-language neurons." But here was the surprise: even when those specialized cells were removed from the analysis, the broader shared neural geometry remained intact. This revealed that translation is not orchestrated by dedicated "dictionary neurons" but emerges from coordinated activity across large neural populations.
Dr. Benjamin Hayden, co-senior author of the study, offered a vivid explanation: the brain is like a piano. Each neuron is a key. When someone hears the same phrase in different languages, it is like playing the same song in different musical keys. Different keys must be depressed, but the relative pattern between successive notes—the melody's shape—remains identical, just transposed. The activity of any single neuron differs between languages, but the pattern across the population stays the same.
To verify these findings, the researchers compared human brain activity to multilingual artificial intelligence language models, including multilingual BERT, and found striking similarities between the geometric patterns. This convergence between biological and artificial neural networks suggests the team has uncovered a fundamental principle of how intelligence organizes meaning.
For the millions of people who navigate multiple languages daily, the research offers a new understanding of how their brains work: not as competing systems struggling for dominance, but as a unified semantic landscape accessed through different linguistic doorways. The work, posted to the bioRxiv preprint server, suggests that fluent bilingualism is not a balancing act—it is a reflection of how the brain has evolved to think about meaning itself.